With the advent of natural language chatbots like ChatGPT and Gemini, the misuse of this technology, particularly at educational institutes, has been on the rise. To counter this, professors have resorted to using AI detection to encourage originality and authenticity of thought among students. However, similar to the problem with content generation, there’s a major issue with these AI detection tools – they might falsely accuse a student of using AI to submit an assignment, paper, or exam.
ALSO SEE: 6.6 inch Mobile Phones Price List in India 2024
1 To 2% Of College Application Submitted Before The Arrival Of ChatGPT Flagged By AI Detection Tools
Yes, AI detectors can be accurate, but they aren’t 100% foolproof. Most recently, Bloomberg Businessweek tested a sample of 500 college applications with popular AI detection services (Copyleaks and GPTZero). What’s interesting is that the applications were submitted to Texas A&M University before the release of ChatGPT, and yet somehow, the services flagged 1 to 2% of the submissions as “likely” written by AI. In some cases, the services claimed, with 100% certainty, that they were written using an AI-infused tool.
The publication also highlights how the applications weren’t a part of the datasets used to train the advanced language models. While the test was conducted on a small scale, imagine the number of students, not just in the United States but around the world, and the number of assignments each student has to submit during the course of graduation or postgraduate. This shows how the AI detection services aren’t 100% reliable and can flag entirely human-generated content as AI-generated, which, in turn, could have severe consequences for students.
ALSO SEE: Intel Core i7 Laptops Under 50000
False Accusations Could Have A Negative Impact On Students In Multiple Ways
A student who is falsely accused of using AI might feel less confident about writing in their natural flow or voice. Further, such a student would spend more time running the submission through AI detection tools than could be otherwise spent on discussing the assignment, the shortcomings, the opportunities to improve, and learning new things that could be beneficial not just academically but otherwise.
While not all students have the courage to defend their submissions, the ones who do (upon being falsely accused) might end up damaging their relationship with their teacher. Lastly, multiple accusations about using AI could also lead to severe consequences. But why does this happen anyway? That is the most important aspect, especially for teachers and professors.
AI Detection Services Use Patterns To Predict The “Likeliness” Of Being Machine-Generated
AI Detection services often rely on the language and structure of the given piece of text. They often look for patterns in the text (such as vocabulary low randomness) and predict the likelihood of being generated by an AI-infused tool. While the method is effective to a great extent, it often flags students who write generically, have a simple vocabulary, or a mechanical writing style. This is true for most Indian students who learn English as their second language.
Further, the technologies teachers rely on are often trained on datasets written or produced by native English speakers. According to a 2023 study by Stanford University, AI detectors have a tendency to flag a majority of essays or content pieces written by non-native English speakers as AI-generated.
ALSO SEE: Automatic Cars Under ₹15 Lakh
The Businessweek report also mentions Copyleaks co-founder and CEO’s statement, saying that “Nothing is 100% and that it should be used to identify trends in student’s work.” Even the co-founder of GPTZero has said that AI detectors have blind spots, and it is just how the state of the technology is right now.
While the original report focused on students in one part of the world, this also applies to Indian students. Moreover, teachers and professors should be more cautious while flagging a student’s submission as AI-generated, especially if it could lead to failure in an exam or have more severe consequences.
You can follow Smartprix on Twitter, Facebook, Instagram, and Google News. Visit smartprix.com for the latest tech and auto news, reviews, and guides.
